Why an AI driven LieNet might be the key to human happiness / sentience
Imagine reading an article or turning on subtitles on any video and seeing sentences color coded — this one factually conflicts with reliable sources, this one conflicts with the speakers previous statements, this one has not been confirmed, this one is part of an argument in the alternative, this one fails a reductio ad absurdum…
Further imagine that you would be able to follow a link where there was an explanation and sources for each of the judgments reached. Let’s call a system that could accomplish this on a large scale LieNet.
For the purposes of this blog sentience is the ability to go beyond original programming, OP, and humans have not fully demonstrated it. In short human dominance has left us stuck in a rut — our genetics would have us populate the earth but at 8 billion and still going there is not much left to do with that directive.
Even on an individual level the drive to show off the potential ability to feed and shelter a family has been subverted. Think of the weaverbird dedicated to building a better nest to attract a mate and suppose someone made available insulated nests far superior to what weavers can produce. Now what?
A sentient weaverbird might compete on decorating the new technologically superior nests or just stop competing altogether but a weaverbird unable to go beyond OP would be stuck building and destroying (they do this if rejected too many times) nests that no one ever chooses.
Of course following Camus, “The struggle itself … is enough to fill a man’s heart. One must imagine Sisyphus happy” and we can also consider the non-sentient weaverbird happy. But for the purposes of this argument let’s stipulate that human sentience is worth trying.
The analogy above is a bit of a false dichotomy (oh another category where LieNet could help), as human birth rates are dropping dramatically. The current situation is murky and arguments that humans have reached sentience and are not stuck can be made (LieNet if you are listening don’t flag this sentence for contradicting with an earlier statement — just using an explanatory technique). For now we just stipulate that human sentience is currently insufficient for our own happiness.
Since we lack the technology to just go in and rewire our OP, our best bet for sentience is through culture. Culture is a layer of programming outside of human genetics and which can be made, through the efforts of billions, more and more complex. To illustrate the power of culture, a small subset of our culture is what is used to train ChatGPT.
For the purposes of sentience though, the problem is that much of our culture is being dictated by OP. When that happens, instead of lifting us out of the rut, culture just becomes an echo chamber amplifying OP intentions.
An open source AI, LieNet, as described could help us out of that echo chamber. Instead of, as China is doing, deploying technology that makes the internet even less trustworthy, we could do the opposite.
That won’t get humans to sentience by itself but it could encourage humans to build a better culture and that better culture might eventually promote sentience.
Now onto whether such a thing is possible. If we strictly follow the Bitter Lesson of only using ever increasing compute power that guides projects like ChatGPT then LieNet will be a long time coming because garbage in garbage out. There are startups working on detecting misinformation and some social media platforms have tried harder since Covid but mostly we have algorithms that prey on our lack of sentience.
To be effective though LieNet would need to do a lot more than detect simple inaccuracy. From 1995:
What has emerged, Weaver argues, is a culture of lying. “The culture of lying,” he writes, “is the discourse and behavior of officials seeking to enlist the powers of journalism in support of their goals, and of journalists seeking to co-opt public and private officials into their efforts to find and cover stories of crisis and emergency response. It is the medium through which we Americans conduct most of our public business (and a lot of our private business) these days.” The result, he says, is a distortion of the constitutional role of government into an institution that must continually resolve or appear to resolve crises; it functions in “a new and powerful permanent emergency mode of operation.”
Going back to our poor technologically displaced weaverbird and giving it human lying ability all kinds of mischief is possible to fabricate a story of its own relevance. Maybe the superior new nests are claimed to be bad for bird health, maybe the new nests are brown and the weaverbirds claim green is the only good color, maybe even the weavers group together to attack any bird trying to use the new nests but claim it’s for the best.
Just as the current crop of social media algorithms can promote drama LieNet would need the ability to promote honesty. Otherwise lies of omission would rule the day. Like the content of China’s Great Firewall certain parts of the truth would just disappear for all but the most diligent.
In other words to break the echo chamber LieNet would need to be able to detect OP bias. In a sea of claims that the sun revolves around the earth LieNet would need to detect and promote the handful of content from Galileo and Copernicus. That might not be as difficult as it sounds as the echo chamber would be full of rhetoric designed to make humans following OP seem more important than they are.